FIG. 13 illustrates a flowchart 1300 that shows a training of the machine learning module 106 for balancing shared resources for an interface in a storage controller, in accordance with certain embodiments;

Control starts at block 1302 in which the use of the storage controller 102 is initiated. The process determines whether a predetermined number of I/O operations have been performed in storage controller. If so (“Yes” branch 1306), then control proceeds to block 1308 in which the process computes the margin of error for each port, and the computed margin of errors are used to perform (at block 1310) back propagation in the machine learning module 106, where the margin of errors are computed based on local queuing, global queuing or a combination of both local and global queuing.

If at block 1304 the predetermined number of I/O operations have not been performed in the storage controller (“No branch 1312), then control returns again to block 1304.

FIG. 14 illustrates a block diagram 1400 that shows the adjustment of weights of a plurality of machine learning modules 1404, 1406 of a plurality of storage controller 1410, 1412 from a central computational device 1414 for load balancing of resources in a storage controller, in accordance with certain embodiments (as shown via reference numeral 1402). In certain embodiments, only the central machine learning module 1416 that executes in the central computational device 1414 performs back propagation and then shares the weight and bias changes with the local machine learning modules 1404, 1406 of the storage controllers 1410, 1412.